The version 0.4 is a merge of heitai3+veela10+fsV12.
Version 1.0 is a combination of heitai4+veela11+fsV12.
By adjusting the tag, 2D and 2.5D styles can be produced. Simultaneous use of three triggering words is necessary for creating a 2.5D image, whilst no triggering words are required for the default 2D animation output.
Creating images exceeding an aspect ratio of 16:9 or 9:16 is discouraged due to the potential for serious distortion.
We recommend the use of Anything VAE and DPM++ 2M SDE Karras Sampling method for image generation.
Resolutions over 768px should be avoided.
Adding 'Bad hands' is recommended to reduce the likelihood of undesirable outcomes. The use of any magnifying plugins for enlargement is not recommended.
Typically, utilizing 'img2img' with 'R-ESRGAN 4x+ Anime6B' provides this model with performance surpassing that of 'Hires. fix'.
Link to the model used for synthesis:
https://civitai.com/models/12606/meinahentai
https://civitai.com/models/68025/veela
https://civitai.com/models/91864/fs-v12
描述:
AMOG was my first model.
It has seen me grow and it is a secondary model dedicated to erasing the AI flavor.
The newer 1.6+ version of webUI recommends using DPM++ 3M Karras and setting the sampling step between 40-60 is more appropriate.
Versions below 1.6 are fine with DPM++ SDE set to 25-50 steps.
HD restoration and graphs can be used with all algorithms except Latent series.
Directly out of the map within 768 × 768.
Figure within 3840×3840, HD repair within 2560×2560.
It is recommended to install vae-ft-mse-840000-ema-pruned.ckpt to achieve better results.
训练词语:
名称: animatedModelsOf_31.safetensors
大小 (KB): 2082643
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success
